

Summary Skills

I am an applied computer vision data scientist who have had experience in a leading a bunch of core python
other data scientists. I have used my leadership and applied deep learning skills to successfully
execute various computer vision projects. Highly skilled in applied deep learning - Computer
Vision. Good understanding of various computer vision techniques like classification, bounding
box classification and regression(object detection),pixel wise prediction(segmentation), numpy
sequence classification with LSTMs (video classification). Good Interpreter of business
requirements into technical terms and vice versa. Excellent in use case identification from
business requirements.
Achievements include execution of various computer vision projects like building character pandas
identification,fashion style guide recommender, Fashion attribute predictor, activity detection,
image search engine.

tensorflow
My work got accepted in AI & Deep Learning Conference | GTC 2018 | NVIDIA.
I am looking for a role where i can leverage my skills and experience for the betterment of the
Organisation and further improvement to my own learning.
I am a regular AI/ML blogger. I am a firm believer in Andre NG's view of democratising AI. I pass caffe
on my knowledge to the community through various blogs.

Responsibilities openCV

To understand the business use cases from clients and convert them into a well defined
problem statement and explain it to the development team. Machine Learning
To identify data sets required to develop predictive models for solving internal and external
business problems
To fill data gap by gathering data , designing annotationportal and conducting data
annotation by human annotators. Deep Learning
To explore data sets and identify data transformation and data quality needs for targeted
applications
To d evelop algorithms and predictive models to derive insights and business value from
data
To provide leadership and mentorship to other members of the team.
Others
Identify and implement use cases which might help the organisation business development
To interpret results and produce actionable business insights that lead to measurable
Spark RDD,DataFrame : Used for
business and consumer experience performance improvements
Data transformation
To Operationalize, publish, and monitor successful models to shape business and data
MapReduce : Used as a data
science strategy
processing engine
To partner with other departments to solve problems and identify trends and opportunities
R : Experimental works mainly on
To define and develop the program for metrics creation, data collection, modeling, and
structured data
reporting the operational performance
To work cross-functionally to define problem statements, collect data, build analytical
models and make recommendations.
To routinely communicate metrics, progresses and other key indicators to leadership.
To lead and support various ad hoc projects, as needed, in support of Organizations’s
Business strategy.

Work experience

Lead Data Scientist - Computer Vision 2017 - 2018

Cogknit Semantics

https://www.visualcv.com/abhaymise
Leads a team of 15 members including data scientists,product engineers and UI engineers.

Data Scientist 2016 - 2017

Cogknit Semantics
First member of the computer vision team in the organisation

Data Engineer 2015 - 2016

Cogknit Semantics

Senior Research Analyst 2013 - 2015

Edureka

Education

Bachelor of Engineering,Computer Science 2009 - 2013

visvesvaraya technological university

Machine Learning by Stanford University on June 15, 2016 - June 15, 2016
Coursera with 96.1%
Stanford
A machine learning course by one of the pioneer of Artificial Intelligence

Deep Learning Specialisation 2017 - 2017

Deeplearning.ai
It carries 5 deep learning courses with focus on Neural Net , CNN , LSTM and Bi-LSTMs

Attended cs231n CNN for visual Recognition course 2016 - 2016

Stanford
Attended and completed all the assignments of the course

Achievement

A Poster on "Domain Adaption of Image Caption Model for Video Descriptions"was accepted
at AI & Deep Learning Conference | GTC 2018 | NVIDIA
Multiple blogs has been identified as best reads across various ML/AI forums

Projects

Men Style Guide creation 2018 - present

Cogknit Semantics
Recommend best matching attire for men given a particular attire. The recommended attire is
created on the fly with an inspiration from an automated style guide. The automated style
guide brings the trend factor by following feeds of the fashion celebrities and accordingly
create the combination for recommendation.
Built with pandas, numpy, tensorflow, keras

Metadata Extraction from a media content 2018 - present

Cogknit Semantics
https://www.visualcv.com/abhaymise
Extract attributes like scene count,scene boundaries of each scene, scene identification ,
Character identification , Character's facial features identification , activity detection ,
objectionable activity flagging , speaker identification and close caption generation across all
scenes. The extracted attributes will be exposed micro service for further business
monetisation.
Built with tensorflow, Kaldi, Keras,OpenCV

Face and Voice based authenticated wallet 2017 - 2018

Cogknit Semantics
An authentication system which can be provide multi modal authentication with signals from
face and voice . The authentication can be attached against any payment system . In our case
we attached it against a payment system which was running over blockchain.

Automated transcript generation from advertisement video for 2017 - 2017

e-retail
Cogknit Semantics
The project aims in solving a business case the e-retailers website should accessible to a
visually impaired person. Aim is to create a transcripts having visual and audio
information about the given video. This transcript would contain complete information
including all actions , emotions , clothing attributes and spoken sentences. This transcript
would be used by the video owners in explaining their products to visually impaired person.
Built with tensorflow, Kaldi, Keras,OpenCV

Scene Retrieval using Tensorflow and Opencv 2016 - 2017

Cogknit Semantics
This system allows you to retrieve and play a video from the scene that you desire.
Built using openCV as means of IO,Inception-v3 CNN encoder from tensorflow to
extract features of the video.

Image search using tensorflow 2016 - 2017

Cogknit Semantics
This system has been trained to recognise categories of the query image and then to retrieve
the most closest match under the identified category as a search result.
Built with tensorflow CNN classifier, encoder and vectorised search system.

Automated image curation system using tensorflow 2016 - 2017

Cogknit Semantics
This system helps in accelerating the cleaning process of the training data of image classifiers.
It brings out all the junks from the training data and provides a system to delete/move from the
categories.
Built with tensorflow CNN feature extractors and vectorised information retrieval
technique.

Auto correction of e-Commerce catalogue using caffe 2016 - 2016

Cogknit Semantics
This system tries to find the mis-placed item under a category in an e-commerce
catalogue.Once it finds the misplaced item it reports it to the admin and then suggest the
correct category under which the product should actually be placed.
Built with caffe CNN classifier.

Automated Attendance system 2017 - 2018

Cogknit Semantics
This marks an employee presence in office by detecting his/her face at the entry of the office.

https://www.visualcv.com/abhaymise
Built with tensorflow CNN classifier and openCV.

Auto tagging of visual contents for indexing using tensorflow 2017 - 2017
Cogknit Semantics
This system automatically generates a textual description of an image.
This also marks the image with multiple tags as per the objects detected in the image.
Captioning system built with CNN-LSTM encoder-decoder system. Meta tag generation
is built using Faster-RCNN network.

Recommendation System for smart learning using numpy 2015 - 2016

Cogknit Semantics
Recommends learning contents to an user based on his and his peer activities.The learning
process of like minded learners is collaborated to suggest new and useful contents to an user.
Collaborative filtering of user and item attributes has been used .

https://www.visualcv.com/abhaymise